Mask Rom

Mask ROM is a read-only memory whose contents are permanently fixed during manufacturing, encoded directly into one of the photomasks used to fabricate the chip, so the stored data is literally built into the silicon and can never be changed afterward. It is the most basic and lowest-cost form of read-only memory for high volumes, and it sits at one extreme of the memory spectrum: data that is written exactly once, at the factory, by the physical structure of the device itself. Understanding mask ROM clarifies what all the other programmable memory types were invented to avoid — the inflexibility of committing data at fabrication time.

The defining characteristic of mask ROM is that its data is set by a photomask, which ties the memory contents to the manufacturing process rather than to any later programming step. Chips are built by patterning layers of material using photomasks that define where features go, and in a mask ROM the presence or absence of a connection at each memory location — which determines whether that bit reads as a one or a zero — is patterned in by a custom mask. Changing the stored data therefore requires making a new mask and fabricating new chips, which is why the data is unchangeable once the part exists: the information is a physical feature of the device, not a stored charge or a settable state.

Data defined in a photomask
the bit pattern is patterned into the chip during fabrication
Consequences of fixing data at the mask
Permanent, unchangeable
altering data needs a new mask and new wafers
Lowest cost per bit at high volume
no per-chip programming, no erase hardware
High up-front cost + lead time
mask tooling and fabrication before any part exists
Best for stable, high-volume data
finished product firmware, character sets, fixed tables

Mask ROM's economics are its reason for existing: at high volume it is the cheapest way to store fixed data, but it carries a steep up-front cost and no flexibility. Because the data is built in by the mask, there is no per-chip programming step and no need for any erase or write circuitry, so the marginal cost per chip is very low when many identical units are made. Against that, creating the custom mask and fabricating the wafers is expensive and slow, and the entire run is useless if the data turns out to be wrong. This makes mask ROM suitable only when the data is both stable and needed in large quantities, where the up-front cost is amortized over many parts.

The inflexibility of mask ROM is precisely the pain point that motivated the invention of field-programmable and erasable memories. If a bug is found in mask-ROM firmware, there is no fix short of discarding the parts and remaking them, and small-volume or still-changing data cannot justify the mask cost at all. This is why the industry developed PROM, which could be programmed once after manufacture, and then EPROM and EEPROM, which could be erased and reprogrammed — each successive technology traded some of mask ROM's low per-unit cost for the flexibility of being able to set or change the data after the chip was built.

Mask ROM remains the right choice for a specific niche: large volumes of truly stable data where cost per unit dominates. Applications like the fixed firmware of a mature, high-volume consumer product, built-in character fonts, or unchanging lookup tables fit it well, because the data will not change over the product's life and the quantities justify the mask investment. Seeing where mask ROM still wins clarifies the general principle that memory choice is a trade among flexibility, cost per bit, and volume, with mask ROM occupying the low-cost, zero-flexibility, high-volume corner.

Read mask ROM as data cast into the chip itself: a bit pattern fixed by a fabrication mask, unchangeable and cheapest-per-unit at high volume, whose very rigidity defined the problem that PROM, EPROM, and EEPROM were each invented to solve.

Take mask ROM further

Ask the copilot about this term, or have our engineers assess it against your process.